def display_uploaded_img(contents, fname, date):
if contents is not None:
original_img, resized_img = parse_image(contents, fname, date)
img = np.expand_dims(resized_img, axis=0)
prediction_array = model.predict(img)
prediction = np.argmax(prediction_array)
children = [
"Your uploaded image: ",
html.Img(className="image", src=original_img),
"Image fed the model: ",
html.Img(className="image", src=create_img(resized_img)),
f"The model thinks this is a {label_mapping[prediction]}",
html.Button(id="clear-button",
children="Remove Uploaded Image",
n_clicks=0),
]
return children
def display_selected_point(clickData):
if not clickData:
raise dash.exceptions.PreventUpdate
idx = clickData["points"][0]["customdata"]
img = np.expand_dims(all_images[idx], axis=0)
prediction_array = model.predict(img)
prediction = np.argmax(prediction_array)
probability = np.round(prediction_array[0, prediction] * 100, 2)
ground_truth = all_labels[idx]
correct = prediction == ground_truth
if correct:
color = "green"
else:
color = "red"
return [
create_img(all_images[idx]),
[
f"prediction: {label_mapping[prediction]} ({probability}% certainty)",
html.Br(),
f"actual: {label_mapping[ground_truth]}",
],
]
def display_selected_point(hoverData):
if not hoverData:
return create_img(train_images[0])
idx = hoverData["points"][0]["customdata"]
return create_img(all_images[idx])
html.Div(
id="prediction",
children=[
"Click on a point to display the Network's prediction",
html.Br(),
html.Br(),
],
),
],
style={"height": "20%"},
),
html.Br(),
html.Img(
id="selected-data-graph",
className="image",
src=create_img(
np.zeros((28, 28))),
),
],
)
],
),
],
),
],
),
],
),
])
@app.callback(
